Heretofore, there has been proposed a technique in which a curvature radius of a curve in a road, limited-speed information, traffic lane information, and the like are acquired from map information, to thereby optimize the speed of a vehicle. In this regard, if a position and a speed as references for optimization are determined in a single uniform way, a possibility arises that the vehicle speed becomes minimum before a preferable position or the start of acceleration is significantly delayed, and thus driving assistance desired by the driver is not performed.
In that regard, it is also conceivable that the optimum speed is learned based on a driving operation by the driver; however, the amount of information related to the driving operation is huge during traveling of the vehicle, so that, in order to store them and to obtain the optimum speed, a large capacity memory and much calculation processing are required.
Further, because a driving operation unwanted by the driver is also learned, there is a case where a driving control is performed with an uncomfortable feeling for the driver.
On the other hand, in Patent Document 1, there is disclosed a traveling control device that calculates a target speed and a target acceleration rate based on a state of the preceding vehicle and map information, to thereby control traveling of the host vehicle.
Meanwhile, a driving assistance system described in Patent Document 2 records, when the host vehicle enters a target area such as a corner, respective sets of feature points corresponding to plural different types, to thereby generate based on the recorded feature points, a variation map in which degrees of variations of the feature points are shown for each of the types. Then, driving assistance is executed so that a feature point having an average value of the degree of variation in the variation map is reproduced.